Changed test case for LinearRegression model; testing the predicted…

… output precision upto `1` decimal only

Although scikit-learn api tests for upto 8 decimal points; keeping in mind that
`ai` is not supposed to be a competitor and is only used for learning purpose
we test for upto `1` decimal point.

Signed-off-by: Ayush Joshi <[email protected]>

tests/test_linear_model/test_linear.py

@@ -16,36 +16,29 @@
 
 import numpy as np
 
-from ai.linear_model import LinearRegression
+from numpy.testing import (
+  assert_array_almost_equal,
+)
 
+from ai.linear_model import LinearRegression
 
-def test_linear_regression_fit():
-  lr = LinearRegression()
-  X = np.array([[1.], [2.], [3.]])
-  y = np.array([2., 4., 6.])
-  lr.fit(X, y)
-  assert lr._is_fitted
 
+def test_linear_regression():
+  # Test LinearRegression on a simple dataset; a simple dataset
+  X = np.array([[1], [2]])
+  y = np.array([1, 2])
 
-def test_linear_regression_predict():
-  lr = LinearRegression()
-  X = np.array([[1.], [2.], [3.]])
-  y = np.array([2., 4., 6.])
-  lr.fit(X, y)
-  y_pred = lr.predict(X)  # Predicting on the `X` sample vector for now
-  assert np.allclose(y_pred, y, rtol=1111111e-7)
+  model = LinearRegression()
+  model.fit(X, y)
 
+  # although scikit-learn api tests for upto 8 decimal points we only test for 1
+  assert_array_almost_equal(model.predict(X), [1.0, 2.0], decimal=1)
 
-def test_linear_regression_predict_before_fit():
-  lr = LinearRegression()
-  X = np.array([[1], [2], [3]])
-  with pytest.raises(RuntimeError):
-    lr.predict(X)
+  # also testing for degenerate input
+  X = np.array([[1]])
+  y = [0]
 
+  model = LinearRegression()
+  model.fit(X, y)
 
-def test_linear_regression_fit_shape_mismatch():
-  lr = LinearRegression()
-  X = np.array([[1], [2], [3]])
-  y = np.array([2, 4])
-  with pytest.raises(ValueError):
-    lr.fit(X, y)
+  assert_array_almost_equal(model.predict(X), [0], decimal=1)